The world’s biggest chipmaker’s chief executive has managed to get President Donald Trump out of his back this year, but has yet to beat investors in his strategy to ride the current trade storm.
Taiwanese semiconductor manufacturer CC Wei announced its $100 million investment in the US on March 3rd. This has silenced Trump’s accusations that Taiwan is “stealing” US semiconductor businesses for now.
Instead, the US President now says that the additional investment from TSMC will bring a “large proportion” of chip manufacturing to US territory.
Investors aren’t very enthusiastic, with TSMC’s stock down more than 17% from its record hit shortly after Trump’s take office in January.
Their concerns are two. Trump can turn TSMC on again if he can’t live up to his high expectations, but the rapid and massive expansion of US manufacturing can reduce the company’s great profit margins.
The contracted chipmaker has little to say how quickly they intend to spend the extra $100 billion and exactly what they intend to spend on. That contrasts with five years ago when he first committed to investing in production capacity in Arizona during Trump’s first term.
At that point, it gave the start date of construction, named the generation of the process technology to be used, and said that commercial production will begin and how many chips it will solve.
Raising the investment plan to $400 billion in late 2022 and $650 billion in April 2024 gave similar levels of detail.
Now, the company says that simply building three more manufacturing plants or fabs on top of the three previously announced facilities will result in two more facilities being built to package multiple chips on one substrate.
“They had already planned to build a lot of fabs here. Did that accelerate those plans? I’m a bit skeptical,” said Peter Hanbury, Bain’s partner, focusing on technology manufacturing and semiconductors. “As the advanced packaging fabs were not on the roadmap, they can be considered a progressive new production.”
There are no steps to the company’s research footprint either. TSMC says the “R&D Team Unit” which has pledged to install it in Arizona rather than developing cutting-edge new technologies, has already committed to gradual adjustments to the mass production process.
“That R&D is one of the big challenges to stay in Taiwan,” Hanbury said. He added, whether it’s TSMC or its rivals Intel or Samsung, “people aren’t moving semiconductor R&D into new geography under almost every scenario.”
“Unless you design next-generation transistor technology in the US, you don’t have leadership in the US,” former Intel chief Pat Gelsinger told the Financial Times last week.
People close to TSMC suggest that the additional $100 billion is merely an estimated price tag for the plan that was featured in the book in the long term anyway, adding that the White House announcement reflects intentions rather than promises.
The company acquired enough land in 2020 to build six fabs. “We’ve always said it was for future expansion,” said someone familiar with the plan. “Now, our first fab is commercially produced and the yields are good, so we can keep watching ahead.”
TSMC declined to comment.
The extent of TSMC’s responsibilities from pressure and threats from the US administration is unknown.
Trump has repeatedly said he wants to impose tariffs on imported semiconductors. This will increase costs for TSMC customers such as Apple and Nvidia, and will likely require manufacturers to share the burden.
Currently, Washington is expected to push back such tip tariffs until it announces “mutual” tariffs on April 2nd. Experts in the delay industry said subcomponents are most likely due to the challenges of virtually challenges.
Trump’s “chip story was overcome by events,” said one US official.
Some view relief as temporary. One senior executive at TSMC said: “He obviously liked the $100 billion figure, but don’t expect that to be enough.”
TSMC’s US expansion remains paler than investments in its Taiwanese home base. There, they say they intend to build 10 or 11 additional fabs. Bernstein analysts estimate that TSMC’s Arizona plant will account for a third of its total revenue by the early 2030s.
“If all of these $165 billion is spent, we could reach capacity. This could amount to around 23-28% of the cutting edge (chip) by the end of 2026,” he added that the percentage “should be smaller than that,” as the build-out takes much longer than the second half of 2026.
“I don’t think that companies around the world can simultaneously expand six fabs, especially in the US,” Lynn said, pointing to potential bottlenecks in equipment and labor.
TSMC needs to pace the US expansion to avoid damaging profitability too much. Bernstein analysts calculate that Arizona Fab (which has now enacted losses) will need to achieve a 40% total margin in the early 2030s and maintain a company-wide margin of over 53% as a forecast.
In the long run, the photos will be much brighter. A TSMC executive said the company needs to expand overseas due to restrictions on how large Taiwanese engineers and pools of land, electricity and water resources can be at home.
Recommended
Experts said the size of the Arizona Operation is equal to the size of TSMC’s “Gigafabs” homeland, and should be able to reach similar efficiency levels once all capacity is commercial production.
But that will be long after Trump took office. The current concern lies in whether TSMC’s big pledge has ended the administration’s push to help revise the manufacturing operations of its struggling rival Intel.
According to industry experts, such plans are “impossible” and the tools and processes of the two chip makers are so different, it could cost more to change Intel Fab than to build something completely new.
However, TSMC client executives believe Washington has bigger plans to win TSMC. “They’re going to bring you back Intel and make them invest a lot. After you build everything, they say you’re monopoly now and you have to sell, we’ll find a buyer,” he said. “I think that’s the endgame.”
